Jet Propulsion Labratory

MEETING SUMMARY - written by George Aumann, AIRS Project Scientist

Note: The numbers in parentheses refer to presentations found below the core text.

The AIRS Science Team met on March 30, 31 and April 1, 2004 in Greenbelt, Maryland. Presentations topics included instrument status, the performance of the data products and planned improvements, data assimilation for weather forecasting, routine and field campaign data product validation, and emerging research products. A number of products, such as the surface air temperature, related to the shelter temperature as opposed to the surface skin temperature, tropospheric and stratospheric water vapor demonstrate that the potential expected from AIRS is being realized (1). All systems on AIRS and AMSU are performing nominally and promise to deliver a seven-year record of the global upwelling radiances with unprecedented accuracy and spectral resolution (2).

Calibration:

The AIRS level 1B absolute radiometric calibration (3) remains validated at the 0.2 K level predicted by the pre-launch calibration. The validation in the 275K to 305K temperature range is based on comparisons with the NCEP RTGSST.Recent field measurements in Antarctica have extended the validation to 250K (4). The first assessment of the long-term stability of the AIRS absolute radiometric calibration relative to the NCEP RTGSST shows no significant trends at the 8mK/year level since September 2002 (5).The spectral calibration of AIRS relative to the upwelling radiance has been better than 4 ppm during the same time (3).The absolute radiometric calibration of the AMSU Instrument is more complicated (6), but good results are being obtained using scan angle dependent bias tuning. Evaluation of AMSU radiometric stability relative to AIRS was initiated in October 2003 (5).

Standard Products and Validation

The excellent radiometric performance of AIRS provides a stable source of data for the generation of standard and research products based on radiative transfer calculations. Refinements of the radiative transfer (7) have been based on over one million matchups of cloud-free AIRS data with ECMWF and dedicated radiosondes. Version 3.1. of the AIRS Standard Products software for clear and cloudy, ocean, land, day, and night is currently available from the GSFC DAAC. Version 3.5 was delivered to NOAA NESDIS as part of the final testing phase and is expected to be installed at the DAAC in October 2004 (9).The internal quality control in version 3.5 identifies 30% of all available day/night non-polar ocean retrievals as high quality. These retrievals meet the 1K/1km accuracy requirement (10).Water vapor profiles show excellent agreement with ECMWF analysis and radio sondes with 20% accuracy in 2 km layers (11). A considerable effort is being expended on developing a consensus on the accuracy of the field measurements as function of the instrumentation (12,13,14,15). Validation of the sea surface temperature and emissivity are making good progress (16, 17) and were the subject of a splinter meeting. A special issue of JGR is planned to report on the validation of AIRS data. The papers are expected to form the core of the first release of the post-launch AIRS Level 2 Validation Report.

The shift from non-polar ocean to non-polar land requires significant refinements in the algorithm (18, 19, 20). The inhomogeneity of land temperature, moisture and surface temperature and emissivity requires reevaluation of some of the basic premises of the retrieval algorithm, surface emissivity validation (21) and refinements in the internal quality control.

Cloud-clearing

While the evaluation of the radiometric calibration is based on cloud-screened ("cloud-free") data, the retrievals are based on cloud-cleared spectra, using the combination of AIRS and AMSU data (20). A number of alternative cloud-clearing approaches are under evaluation (23). Cloud-cleared radiances are a standard level 2 product. While less than 1% of the day/night non-polar ocean data are cloud-free at the detector noise level, version 3.5 retrieval returns high quality cloud-cleared spectra for 30% of the day/night non-polar ocean data. The fact that retrievals using these spectra produce retrievals with 1K/1km accurate constitutes a form of validation. In addition, cloud cleared radiances are screened using spectral tests (22). Cloud-clearing results in a measurable increase in the noise compared to cloud-free spectra, which is folded into the 1K/1km accuracy assessment. 99% of the cloud-cleared data which pass the retrieval quality control also pass the spectral test.

Data Assimilation

AIRS Level 1B (radiances) data have been distributed to the NWC through NOAA/NESDIS since December 2002 (24). Data is being assimilated at the GSFC/DAO (25) as temperature and moisture profiles. This is done retrospective using January 2003 data processed with version 3.5 software. NCEP (26), ECMWF (27) and UKMetOffice (28) assimilate radiances. Assimilation at NCEP is in the parallel run real-time system; assimilation at ECMWF is in the operational system; assimilation at UKMeto is expected to go into the operational system in May 2004. All NWC report a small, but significant forecast skill improvement with AIRS data in the southern hemisphere. The operational system at ECMWF now includes a retrieval for CO2.

Assimilation at NCEP, ECMWF and UKMetOffice is currently limited to essentially cloud-free radiances. With version 3.5, temperature and moisture profiles and cloud-cleared radiances are available for distribution to the NWC. The use of quality screened cloud-cleared spectra increases the amount of data available for direct assimilation by a factor of 30.

Climate Research

AIRS data supports the NOAA data requirements for weather forecasting and NASA s interest in climate research. Such research requires very high quality data, which are well documented, have no hidden biases, and are readily accessible for long term trending. Initial evaluations show that data access is already becoming a problem after 18 months of data. A number of AIRS Science Team members are experimenting with data subsetting. The analysis of trends from cloud free day/night ocean spectra shows great promise (29). Without loss of accuracy, the AIRS data can be compacted by a factor of two compared to the current storage at the DAAC (30). More aggressive, but still lossless compression can result in a factor of four reductions in the data volume (31). Much higher levels of compression can be achieved by saving only the most significant principle components (32) or by creating data subsets. The daily, weekly and or monthly mapping of level 2 products on a global grid (Level 3 data products) is another form of information preserving compression, which is under development (33). Researchers who are planning to get AIRS data from the GSFC/DAAC have expressed considerable interest in Level 3 products (34).

Research Products

The wealth of information in the infrared spectrum is starting to be exploited in a number of research products, which may be upgraded to standard products when the demonstrated accuracy reaches a stable and competitive level. The accuracy of the retrieval of Ozone with AIRS already appears to be comparable with TOMS (35). The algorithm is already integrated into the product generation system a the DAAC. Offline experimental algorithms for CO (36), aerosol, dust and SO2 (37) are also very promising. The out-going long wave radiation (OLR) based on cloud-cleared radiances is already a recognized research product. An effort is starting (38) to evaluate the far-infrared properties of the Earth radiation budget derived from AIRS and CERES.

The next AIRS Science Team meeting will be from 13-15 July 2004 in the Pasadena, California, area.

Details on the referenced presentations from the AIRS Science Team Meeting can be downloaded by clicking on links at the end of the presentation summaries below.

(1) Moustafa Chahine (JPL), AIRS team leader, announced that the AIRS Standard Products from non-polar day and night ocean data meet the 1K/1km science requirement. The development of monthly geophysical products show great promise for the refinement of climate models. The wealth of information in the infrared spectrum is starting to be exploited in a number of research products, which may be upgraded to standard products, when their global validation matures and the accuracy reaches a competive level.

(2) Tom Pagano (JPL) reviewed the AIRS, AMSU and HSB Operations Status. He discussed AIRS/AMSU/HSB data products, their relationship to the "Earth Science Questions", the schedule for their global validation and the validation status. AIRS and AMSU level 1b and level 2 data are globally distributed in support of weather forecasting. The development of research products, support for climate research and AIRS Education and Public Outreach efforts are increasing. [download presentation]

(3) Steve Gaiser (JPL) reported that the AIRS IR calibration algorithms are performing extremely well. Radiance validation using the aircraft campaigns (SHIS) and routine monitoring relative to the NCEP RTG SST are consistent with the 0.2K absolute accuracy predicted based on the pre-launch calibration. The radiometric performance has been extremely stable. The sea surface skin temperature derived from 2616cm-1 under cloud-free conditions is stable relative to the NCEP Real Time Global Sea Surface Temperature (RTGSST) at the 8mK per year level. The stability of the spectral calibration is monitored routinely relative to the upwelling spectral radiance and is stable at the 4 ppm level, considerably better than the required 8ppm level. All IR calibration algorithm improvements currently under consideration would result in changes of less than 0.1K in brightness temperatures. [download presentation]

(4) V. Walden (U. Idaho) reported on AIRS level 1b evaluation using surface measurements as Dome Concordia in Antarctica during January 2004. The validation used the Polar Atmospheric Emitted Radiance Interferometer (PAERI) and an Infrared Radiometric Thermometer (IRT) during the six daily overpasses of the EOS Aqua. The PAERI measured spectral IR radiances of sky and surface are calibrated to within 0.05 K at -20 C (253 K) . The Infrared Radiometric Thermometer (IRT) is mounted on a snowmobile. Both units were operated also from a 32-meter tower and were frequently inter-calibrated to the SSEC BB in the field. Most preliminary comparisons between the measurements of AIRS in the 8-12 micron window channels agree within +/- 0.5 K. A few data sets have larger differences that still need to be resolved. [download presentation]

(5) George Aumann (JPL) presented AIRS and AMSU Level 1B trend analysis based on cloud-screened data. Establishing absolute stability is key to generation of data in support of climate data records.The AIRS radiances are based on the NIST traceable onboard calibration blackbody. These radiances are tied via the 2616 cm-1 window channel to the NCEP RTGSST. Analysis of 15 months of AIRS 2616 cm-1 based sea surface temperature measurements relative to the RTGSST between September 2002 and March 2004 show radiometric stability at the 8mK level. This establishes the radiometric stability for all AIRS channels through the common onboard blackbody and space view. The radiometric stability of AMSU was tested by comparing the raw brightness temperatures for the AIRS 2388cm-1 channel and AMSU channel 5 at 53 GHz. The two channel have closely matching weighting functions, but the opacity at 2388 cm-1 is due to CO2, while the opacity at 53GHz is due to oxygen. For the six months between October 2003 and March 2004 the raw brightness temperatures differed by only 1.8K with 0.5K standard deviation. The design of AMSU is such that stability in one of its channels does not establish stability at a comparable level in all fifteen channels. [download presentation]

(6) B. Lambrigtsen (JPL) reported on the status of the AMSU calibration. A more accurate calibration algorithm is being developed that is expected to improve the radiometric accuracy by up to 1 K. This approach takes into account that the radiometric zero point, which is currently derived from the space view, is contaminated by Earth's radiometric background, which varies along the orbit. The moon's effect on space-view calibration observations will also be modeled and accounted for, which will eliminate small data gaps when the moon enters into the cold-calibration FOV. Recent pointing analysis results were also presented, based on observations of coastline crossings while in nadir stare mode.Pitch errors are on the order of 10% of the IFOV. [download presentation]

(7) Larrabee Strow (UMBC) presented details of the Radiative Transfer Algorithm (RTA)upgrades. The RTA is the heart of the physical retrieval algorithm. The RTA has been refined based on more than one year of RS-90 sonde launches during cloud-free AIRS over flights of the ARM-TWP. AIRS spectral cloud screening and visible channels were uses to insure that the RTA validation data sets was in fact cloud-free.This is particularly important the 12 micron region water vapor continuum was modified. Global maps of window region biases suggest that remaining RTA errors are complicated, with some components due to the water continuum, other components due to incorrect surface emissivity (near 800 cm-1), and some possibly due to inaccurate models for reflected thermal radiation.

(8) Steve Friedman (JPL) presented details of the PGE software plan for the next year, leading to the delivery of Version 4.0 to the GSFC DAAC, which currently runs version 3.1. With the delivery scheduled for October 2004, the software development will close in July 2004 to allow for adequate testing and documentation. The delivery to the DAAC will include updated Algorithm Theoretical Basis Documents, and the Validation Report. [download presentation]

(9) E. Manning (JPL) discussed improvements in the retrieval algorithm incorporated in V3.5. This version, which was delivered to NOAA/NESDIS for evaluation, was extensively validated for ocean within 50 degree latitude. The main improvement is due to much tighter quality control, which is summarized in 16 bit retrieval QA flag. Less than 1% of the retrievals result in surface temperatures outliers, defined as abs(RTGSST-TSurf)>3K. This version includes the experimental Ozone retrieval, but the experimental CH4 retrieval was turned off. [download presentation]

(10) S. Y. Lee (JPL) discussed the evaluation of Version 3.5 using retrievals from 12 "focus days".Focus days are typically every 48th EOS Aqua day between September 6, 2002 and December 5, 2003. About 30% of all retrievals attempted over ocean between 50S and 50N result in an acceptable retrieval. This produces very good global ocean coverage. The 30% yield has to be compared to the 1% yield if retrievals only from cloud-free spectra were acceptable. The rms temperature difference between ECMWF and AIRS is just below 1K, the humidity profile difference is about 20%. [download presentation]

(11) Eric J. Fetzer (JPL) reported on the overall progress of level 2 product validation and on his work with Annmarie Eldering on validating water vapor retrievals. Analysis of time series of retrieved and matching radiosonde temperature and moisture profiles over ocean along the US west coast and Hawaii show that the retrievals in version 3.5 meet 15% / 2 km RMS at the Tropical Western Pacific ARM site at up to the 300 mb pressure level. RAOBs are not reliable above 300mb.He found that "retrieval type zero retrievals are biased toward dry conditions. The results over the eastern Pacific are similar to Larry McMillin s conclusions based on global RAOB matchups using NOAA 17 (see presentation 15). [download presentation]

(12) David Whiteman (GSFC) presented motivations for and preliminary results from the AIRS Water Vapor Experiment-Ground (AWEX-G) held October 27 - November 19, 2003 at the Department of Energy's Southern Great Plains test site in northern Oklahoma. One of the goals of AWEX were the characterization of many of the water vapor measurement techniques in use for AIRS validation and the reconfirmation of the accuracy of the permanently sited DOE water vapor profilers: the Raman lidar and the RS-90 radiosondes. The preliminary results presented included a characterization of Vaisala RS-80 and RS-92 sensors with respect to the University of Colorado Cryogenic Frostpoint Hygrometer. This study indicated a significant moist bias of the RS-80 with respect to the CU-CFH but generally good agreement between the RS92 and CU-CFH. An excellent AIRS comparison case from November 19, characterized by uniform water vapor conditions and very clear conditions, was also shown and will continue to get very careful analysis. Characterization of all the AWEX sensors focusing in particular on upper tropospheric performance will be completed by the end of 2004. David Whiteman's collaborators in this study were Belay Demoz, Larry Miloshevich, Barry Lesht, Holger V mel, Frank Schmidlin, Zhien Wang, Prentiss Moore, Al Beebe, Felicita Russo, Igor Veselovskii, Scott Hannon and Dave Tobin. [download presentation]

Holger Voemel (NOAA/CMDL) reported on results with the Cryogenic Frost point Hygrometer (CFH) measurements during AWEX and previous RS90 comparisons. He concluded that CFH measures on the 5% accuracy level. Snow White, another frost point hygrometer, has no bias issue, but has some performance limitations and the RS-90 has a 5-10% dry bias, possibly due to contamination. May have a slight dry bias 5-10%. VoemelSondes.ppt">[download presentation]

(14) R. Knuteson (U.Wisconsin) reported on Dave Tobin's progress (who was unable to attend) in "ARM Site Atmospheric State Best Estimates for Aqua: Retrieval, Radiance, and Forward Model Validation".Various improvements in the best estimate validation products have been added, including better cloud detection, surface characterization, and uncertainty estimates.Analyses of the best estimate upper level water vapor accuracy were presented. Validation efforts include use of AIRS and MODIS radiances in the clear sky forward model validation, AIRS cloud-cleared radiance validation, and AIRS and MODIS temperature and water vapor profile retrieval validation. [download presentation]

(15) L. McMillin (NOAA/NESDIS) reported on the validation of AIRS water vapor using RAOBs and ground-based GPS with Jim Yoe. The most significant lesson learned was that at the accuracy of AIRS the three types of RAOBs flown in the U.S. have significantly different patterns. Since the U.S. is dedicated to flying more than one radiosonde type, inclusion of a GPS measurement at each site could greatly reduce the differences and is highly recommended. The Vaisala RS-80 currently flown in the U.S. shows good agreement with the GPS. Sippican sondes are not consistent at upper atmospheric levels.

(16) D. Hagan (JPL) reported on the validation of the ocean emissivity and sea surface temperature (SST). A comparison of sea surface emissivity for versions 3.1 and 3.5 demonstrated that V3.5 produces more realistic ocean surface emissivity than V3.1. In both versions, emissivity decreases with increasing view angle, but the decrease is greater than expected based on model calculations (e.g., 2-3% decrease in emissivity, whereas model shows 1-2% decrease for view angles >40 degrees and global mean wind speed of 5 m/s). More SST outliers were observed at large view angles and near cloud boundaries in the final retrieved product compared to the intermediate regression product. Simple tests showed that removing the scan angle dependence of emissivity in the retrieval algorithm, for both day and night, could reduce SST outliers in the final product, suggesting an amplification of errors in the cloud-clearing stage. A splinter session was proposed to address emissivity issues and possible solutions. [download presentation]

(17) Nick Nalli (NOAA/NESDIS) reported on results from the 2004 Aerosol and Ocean Science Expedition (AEROSE), a 4-week oceanographic research cruise onboard the NOAA Ship Ronald H. Brown (RHB) in the tropical North Atlantic Ocean, where he served as the Principal Investigator for remote sensing of SST and aerosols. AEROSE provided a unique complement of marine aerosol data during 2 significant Saharan dust events. Data collected from the expedition will be used to address key questions pertinent to AIRS validation under elevated aerosol conditions. Shipboard radiometric data pertinent to studying the impact of tropospheric dust aerosol upon AIRS spectra and retrievals include the Marine Atmospheric Emitted Radiance Interferometer (M-AERI) (U. Miami), the Calibrated Infrared In situ Measurement System (CIRIMS) (UW/APL) and Microtops handheld sun photometers (Howard Univ.). Vaisala RS80/90 radiosondes were launched ~3 hourly including Aqua overpass times, and standard meteorological data and ocean surface temperatures were acquired by the RHB observing platform. Cross comparisons of the ship radiometric (M-AERI, CIRIMS) and in situ temperatures have been examined for self-consistency and aerosol impact. Investigations into the AIRS quasi-specular reflectance/emissivity models will be conducted using M-AERI radiance spectra and retrieved skin SST and emissivity, along with Vaisala radiosondes. [download presentation]

(18) P. Rosenkranz (MIT) discussed microwave algorithm enhancements for version 3.5. In the version 3.5 retrieval, some residual obs-calc in the microwave channels may be due to microwave tuning or to differences in the calculation of emissivity between the microwave-only and final stages. With respect to biases in estimation of ocean-surface parameters, however, the tuning seems to be more consistent with the revised forward model than the earlier model. Some of the errors in the liquid water product may be coming from warm rain, although at present it is not calibrated for that condition.

Comparing retrievals with and without HSB shows that the failure of HSB in February 2003 had a negative impact on the quality control of the microwave retrievals, and the standard output for some cases changes from type zero (best) to higher type numbers. [download presentation]

(19) J. Susskind (GSFC) discussed the L2 Retrieval algorithm status and enhancement plans in preparation for the transition from non-polar ocean to non-polar land data. Final algorithm updates for version 3.5 included new channel lists for temperature, surface, water, cloud-clearing, and ozone channels; new tuning error estimates for the microwave channels; implementation of the Ts-final minus Ts-initial rejection test; updates of existing rejection thresholds; and limiting the liquid water error estimate in very wet cases. The unexpected scan angle dependence in the sea surface temperature appears to be related to surface reflectivity. An empirical correction to the Lambertian assumption has been implemented.Emissivity and reflectivity issues will be more complicated for the non-polar land retrievals currently under development. [download presentation]

(20) Chris Barnet (NOAA/NESDIS) presented a status and anticipated enhancements related to cloud-clearing and retrieval quality control. The cloud-clearing algorithm in versions 3.1, 3.5 and 4.0 depends heavily on the AMSU channels. Three experiments to test alternate methods of cloud-clearing were discussed, which partially or totally eliminate the reliance on the AMSU. Utilizing temperature retrieval residuals for L2 QA and removing some of the empirical rejection criteria made significant improvements in the lowest 2 km of the atmosphere. [download presentation]

(21) R. Knuteson (U. Wisconsin) reported on land surface characterization using the combination of the high spectral resolution AIRS and moderate spatial resolution MODIS observations from the EOS Aqua. This work was selected as part of the EOS recompetition and addresses the following two important goals: (1) better utilization of satellite sounding data over land and (2) study of the evolution of land surface characteristics. The objectives are to compare the AIRS surface temperature and infrared emissivity products to validation data and to MODIS and ASTER land surface products.Maps of the spectral land surface infrared emissivity will be created for geographic regions of climatic importance. [download presentation]

(22) E. Fishbein (JPL) discussed the quality control of Cloud-Cleared IR Radiances over ocean. The AIRS level 2 algorithm uses the combination of AIRS and AMSU data to eliminate the effects of clouds. The resulting spectrum is then quality tested using spectral tests and retrieval QA.

a) Spectral tests: If a spectrum has been successfully "cloud-cleared", then its spectral characteristics should be indistinguishable from a cloud-free spectrum. The evaluation is based on the comparison of surface temperature in different spectral windows (2616 cm-1, 1231 cm-1, 943 cm-1 and 790 cm-1), and the comparison of the surface air temperature (from the 2388 cm-1, 2390 cm-1, 2392 cm-1 channels) with the surface skin temperature from the 1231 cm-1 and 1228 cm-1 channels.

b) Retrieval tests: High quality cloud-cleared spectra produce retrieval residuals, which fit into an empirically determined QA parameter space. 99% of the retrievals, which pass the retrieval, QA also pass the spectral tests.

There is a measurable increase in the noise and the character of the noise (noise covariance) in cloud-cleared data, which is reflected in the retrieval accuracy. [download presentation]

(23) David Staelin (MIT) presented work by Chuck Cho on Stochastic Cloud-Clearing, which is based on the combination of AIRS channels and AMSU channels 5 and higher.

Preliminary results concerning improved scan angle corrections and the use of a larger training set were reported, together with comparisons of stochastic sea surface skin temperature retrievals with AIRS derived and model based values. Subsequently, the reported bias between the stochastic and model based SST retrievals was reduced significantly by training the estimator primarily with clear pixels.

(24) Walter Wolf (NOAA/NESDIS) summarized the data that are currently being distributed from the NOAA AIRS Near Real Time system to the Numerical Weather Centers in BUFR format. Routine data distribution started in October 2002.88% of the data is currently processed and distributed within the required three hour period. Plans are to start the distribution of cloud-cleared data and level 2 data to the operational user community, as soon as a suitable quality flag can be included. [download presentation]

(25) Robert Atlas (DAO/GSFC) reported on assimilation experiments to assess the impact of AIRS retrievals using both the FVSSI and FVDAS data assimilation systems. Unlike the other NWC, data at the DAO are assimilated using temperature and moisture products (Level 2) derived from AIRS and AMSU data under clear and cloudy conditions. The test used all data from Jan. 2003. In the Southern Hemisphere, there is a significant impact on analyses, and both anomaly correlations and objective cyclone tracking show an overall positive impact of AIRS temperatures on forecast accuracy. In the Northern Hemisphere, the impact is smaller, however on occasion significant impacts occur. Most of the impact of AIRS temperatures comes from high latitude data (poleward of 40 degrees latitude), and both clear and partially cloudy data contribute to the impact. Ongoing and near future work include: evaluating the latest release of AIRS Team retrievals and the impact of AIRS moisture profiles, and comparing radiance and retrieval assimilation. [download presentation]

(26) Stephen Lord (NOAA NCEP) reported on the progress with AIRS data assimilation into the NCEP GCM. Currently the AIRS data are assimilated in experimental "parallel runs", where results can be compared directly to the results of the operational runs. AQUA data processing will become operational at NCEP on April 6, 2004. Use of the AMSU (Advanced Microwave Sounding Unit) instrument aboard AQUA has shown positive forecast impacts in the Southern Hemisphere, even when NOAA AMSU and HIRS instrument data are used. Even with the use of AIRS data limited to a small fraction of cloud free data, AIRS data give small positive impacts in some forecast scores, primarily in the Southern Hemisphere.In the future, measurements from AIRS, AMSU-A, AMSU-B and HIRS will be combined to produce a cloud top and cloud coverage analysis. In addition, window channel radiances will be used within the NCEP global analysis, together with in-situ buoy and ship data, to produce a multi-satellite sea surface temperature analysis. NCEP plans to expand the use of AIRS data sets using reconstructed and cloud-cleared radiances. The AIRS Data Assimilation Team at NCEP includes J. Derber, R. Treadon, D. Kleist and P. van Delst.

(27) T. McNally (ECMWF) reported on the status of assimilation of AIRS radiance data at ECMWF. The forecast of AIRS is slightly positive in the Southern Hemisphere. The current effort focuses on increasing the yield of sounding data in cloudy areas through the utilization of partially cloudy IR data. By introducing clouds as a sink variable within the analysis we hope to explicitly take into account the contribution of the clouds to the measured radiance and use more channels down to the cloud top. He also mentioned that the experimental CO2 estimation is now included in the routine operations and noted that AIRS data force significant deviations from the CO2 background values. AIRS data assimilation at ECMWF is in collaboration with Phil Watts, Marco Matricardi, Jonathan Smith, Jean-Noel Thepaut, Richard Engelen and G. Kelly.

(28) Andrew Collard (UK Met Office), in collaboration with Roger Saunders, James Cameron, John Eyre and Lisa Horrocks, has completed AIRS assimilation trials for Dec 2002-Jan 2003 and Jul 2003.Both show AIRS provides a ~0.5% increase in NWP Index scores. Assimilation of AIRS observations into the global model is currently scheduled to become operational on 25th May 2004. This will be the first change to the operational global assimilation system after the switch to the new NEC SX-6 supercomputer in April. It has been noted that the addition of the Aqua AIRS to the Met Office s two-ATOVS (NOAA-15 and 16) baseline results in an impact on forecast accuracy that is greater than that achieved through the addition of either the NOAA-17 AMSU-A or the Aqua AMSU-A after thinning, implying that AIRS alone and ATOVS impacts on the forecast are similar. This is consistent with work at ECMWF and is very encouraging considering the conservative nature of this initial implementation. [download presentation]

(29) George Aumann (JPL) presented considerations related to the use of the AIRS data in support of climate studies. AIRS currently accumulates data at the rate of 11 TBytes/year.A number of data subsetting approaches are currently under evaluation by science team members to reduce the data volume to about 1 TByte/year or less. As an example of what can be done by saving only 100 AIRS/AMSU and VIS channels for cloud free ocean footprints as 5 MB/day files he showed trends from September 2002 through March 2004. The size of this data set is only 8 GB/year, compared to 11 TBytes/year, compression by more than a factor of one thousand. [download presentation]

(30) Evan Manning (JPL) discussed data utilization issues related to the size of AIRS data products. IR Level 1B files from AIRS alone are producing 121 MB per granule, 29 GB per day, and 11 TB per year. Vis/NIR and Level-2 Cloud-Cleared Radiances add another 2 TBytes/year. This makes it difficult for users to order and store as much data as they would like. The large data volume also makes it more expensive to reprocess the complete data for a uniform climate data quality record. One option is compact the L1b data format to a two byte integer format, compared to the current 4 byte format. This saves a factor of two directly. An additional factor of up to two may be achieved by compression using bzip (or equivalent). He recommended scaling of AIRS Level-1B IR radiances with factor of to NeN, i.e. in a factor four multiple of the signal-to-noise ratio. He recommended not to reorder radiances or implement compression in any AIRS product because the additional size savings does not justify the additional complexity. This reduces the data volume to about 6TB/year. Data users (and GSFC DAAC) should be encouraged to implement external compression or internal compression (using HDF tool hrepack) to save on data transmission bandwidth. [download presentation]

(31) Larry McMillin (NOOA/NESDIS) reported on the evaluation of lossless data compression for AIRS Level 1B data. Using a combination of regression and Huffman coding can produce a 6 bit per channel (average length) file with lossless compression. This could be reduced more by using more codes.In this case, one code was used for all channels. The inclusion of the bad channels in one code table increased the length needed to store the good ones. Using a separate code table for the good channels will solve this.The accuracy was preserved to 0.01K.This is more than is needed to preserve the radiances at the noise/digitizing level of the original data.

(32) Mitch Goldberg (NOAA/NESDIS) reported on AIRS Regression Update and Data Compression Studies. Representation of the AIRS spectra as the first one hundered principal components would reduce the AIRS data volume by a factor of 23 without the data subsetting.

(33) Stephanie Granger (JPL) presented a status report of AIRS Level 3 product development. Level 3 Products are designed to make AIRS data more accessible to a broad range of users by offering moderate size data sets (120 MBytes per day). Level 3 products are derived from Level 2 product, which mapped on the global grid. This makes the data volume much lower than Level 2 or Level 1 (A or B) files. The AIRS Project has developed science requirements for their Level 3 data product and has determined the need for two types of Level 3 products, a conventional product based on means, standard deviations & counts, and an enhanced product that consists of statistical summaries of various parameters based on the distributions of the data within a grid cell. The development of these Level 3 products follows the MISR Joint Products model. The project has generated prototype conventional products and produced a CD containing the prototype products to encourage feedback from the community. Level 3 prototype CDs were distributed to the AIRS Science Team at the meeting. The CD is in html format and contains data description, documentation, a sample reader and a link back to the AIRS feedback forum page to capture comments, questions and suggestions in addition to the data. The software development cycle is underway for development of the conventional Level 3 products with a planned delivery to the DAAC as part of the AIRS v4.0 delivery in Fall 2004. The enhanced AIRS Level 3 product will be a part of the AIRS v5.0 delivery in late 2005. [download presentation]

(34) Gregory Leptoukh (GSFC) presented an overview of the AIRS status at the GES DISC/DAAC. AIRS reprocessing has been moving ahead of schedule. Distribution from the data pool by far exceeds distribution from the archive; therefore, the GES DISC will increase the AIRS allocation. AIRS users have expressed strong interest for Level 3 products, smaller granule sizes, and subsetting options. In response to user concerns, the GES DISC has developed various tools and services for processing, visualization and analysis. Various options for processing "next-to-the-data" were presented. Closer partnership between AIRS scientists/users and the GES DAAC was proposed for customized services development. [download presentation]

(35) Bjorn Lambrigtsen presented work by W. Irion (JPL) on the comparison between AIRS and TOMS Ozone retrievals. Unlike TOMS, which retrieves Ozone only in sun light, the AIRS retrieval in the 10 micron area works day and night. A new set of ozone channels produces a significantly improved agreement with TOMS under v3.2.7 processing for the 9/6/2002 focus day. (AIRS - TOMS) / TOMS median difference (1 sigma) over land is 4.5 8%,over ocean 2.8 6%. However, there are some regional patterns in the AIRS-TOMS, which need to be followed up. Comparison of tropospheric and lower stratospheric retrievals with ozone sondes (in collaboration with Mike Newchurch of UAH) is continuing. The current validation data set includes dedicated ozone sondes from Huntsville and Chesapeake Bay Lighthouse, plus vicarious sonde/Dobson/Brewer/lidar measurements from 143 stations. An evaluation of polar ozone retrievals during polar night, with studies of the vortex morphology entering springtime daylight, is planned. [download presentation]

(36) Larrabee Strow (UMBC) reported on dust and aerosol observations with AIRS and experimental retrievals. Infrared forcing of dust is interesting in it s own right. AIRS spectra of dust allow retrieval of optical depths and some information on particle size. He showed one of many monthly global maps of dust over oceans (using a database of clear-screened "clear" data). The focus is now shifting to more detailed studies of dust events to validate dust optical properties retrievals and to develop dust detections flags.

For the analysis of dust and aerosols in the AIRS data he has developed the SARTA Program, which integrates the RTA and scattering. SARTA appears to return reasonable results for optical depth and particle size. The need for different refractive indices for different regions/storms is not established. So far, nothing stands out (Gobi dust looks like Sahara dust). Afterrefining the methodology on cloud-filtered spectra the plan is to transition to cloud-cleared spectra. Various approaches for estimating the aerosol altitude are being evaluated. A detailed comparison of AIRS derived cirrus and dust optical property retrievals with MODIS retrievals is in preparation for publication.

(37) W. McMillan (UMBC) reported on the global retrieval of Carbon Monoxide from AIRS data. AIRS CO retrieval validation is underway using flask measurements. Retrievals from AIRS are obtained for 80% of globe every day. A sequence of monthly global images shows larges scale CO transport at various altitudes around the globe. 79% of AIRS CO retrievals agree within ~20% with MOPITT values (~ 6000 comparisons/day). AIRS CO retrievals contain vertical information for CO profiling, but this has not been fully explored. The current accuracy estimate is better than 16%, but more validation is needed. INTEX-NA and NAVEX present validation opportunities. His collaborators are C. Barnet (NOAA/NESDIS), L. L. Strow, L. C. Sparling, M. L. McCourt, E. Maddy, S. Datta (UMBC) and P. Novelli (NOAA/CMDL).

1000. M.Mlynczak (NASA LARC) presented plans for the evaluation of the far-infrared properties of the Earth radiation budget derived from AIRS and CERES. This investigation was selected as part of the EOS re-competition and will focus on using AIRS and CERES data to isolate the infrared emission from the Earth and its atmosphere at wavelengths longer than 15 micrometers, i.e., the far-infrared. This spectral region is important because it contains up to half of the radiation emitted by the Earth system and maintains much of the natural greenhouse effect that makes the planet habitable. It has rarely been observed directly. The measurements on the AQUA Satellite offer the only opportunity from space to isolate the radiation from this spectral region and thereby assess the validity of existing radiative transfer models and climate models. [download presentation]